The best trader has the best learning loop.
A loop without a score is just a habit. So: how do you score a forecast?
1 — Scored on aesthetics
Walk into any large organisation. The forecast is a sixty-page deck, clean charts, a Big Four logo, and language that can't be pinned down.
How is it scored? Not by what happens next.
Thick deck. Expensive consultant. Calm language. Therefore the forecast is “good”.
Nobody opens it again. When reality diverges, nobody is graded.
The score happens at the moment of commission. By the time reality arrives, it's already over.
2 — Scored on P&L
Trading floors don't run on decks. They run on books.
A trader's forecast is their position. Their score is their P&L.
If they're right, they get paid. If they're wrong, they lose money. No logo to hide behind. No post-hoc narrative.
P&L mixes skill, risk, and market structure. You can make money and be wrong. You can be right and lose money.
And it only works where markets exist. We have millions of options priced to the second, yet almost nothing on the questions that actually matter. No market for whether a project will blow out. No market for what demand looks like in five years. No market for whether a policy will work.
3 — Scored on Brier
Forecasting competitions took a different path.
Philip Tetlock and others use proper scoring rules like the Brier score.
If you say 70%, it should happen 70% of the time. Say 99% and be wrong, you get punished hard. Saying 50% on everything gets you nowhere.
It's a step up. No market required. Rewards calibration. Comparable across forecasters.
But it has its own failure mode. Brier incentivises smoothness. It punishes conviction. There's no cost to being slightly right all the time. Big, high-impact calls get diluted by volume. And the questions are set by someone else.
What OracleBook does
Take the discipline from markets. Take the rigour from scoring rules. Drop everything else.
A forecast is a number, on the record. Forecasters disagree by committing to different numbers. Resolution comes from canonical data — BOM, AEMO, official prints. Score is calibration plus track record, updated continuously and visible over time.
If you don't like your score, you have to change your forecast. Not your narrative.
The best trader has the best learning loop.
The loop only works if the score tracks reality. Not the deck. Not the logo. Not the story. Just whether you were right, and whether you were willing to say it in advance.